I build AI systems that survive production.

• 

  Enterprise Knowledge Graph

  Agentic pipelines that turn scattered public data on European enterprises into a graph you can actually query.

  2026Agentic AIKnowledge GraphsNLP
• 

  Document Intelligence Suite

  Three parsers reading invoices, emails, and timesheets past 90% field accuracy — including the ones that arrive as photographs of paper.

  2026LLMsRAGDocument AI

• Sales Lead Proposal System

  A lead qualification and proposal engine that predicts purchase intent using semantic similarity rating rather than asking an LLM to score leads directly.

  2025LLMsEmbeddingsSemantic Similarity
• 

  Tunnel & Highway Incident Detection

  Real-time vision for road incidents across Turkish highways and tunnels, running on edge hardware because a datacenter round-trip was too slow.

  2021Computer VisionOpenVINOEdge Compute

Most machine learning work dies in a notebook. My job has been the other part: the async services, the retries, the evaluation harness, and the deployment that keeps a model useful on the thousandth request instead of the first.

Lately that means LLM agents — multi-step tool use, retrieval over knowledge graphs, and the unglamorous plumbing that makes non-deterministic systems dependable enough to put in front of a business process. Before that it was computer vision on edge hardware, where being wrong at 60 frames per second has consequences you can measure in ambulances.

I like problems where the interesting constraint isn't model accuracy but everything around it — latency budgets, messy inputs, and the gap between a benchmark and a Tuesday afternoon.

1. AI Solutions Engineer

   Oct 2021 – Present

   innoscripta SE · Germany · Hybrid

   • Built the agentic extraction pipeline behind a knowledge graph of European enterprise data — LLM agents deciding what to pull, NLP models resolving entities, everything landing somewhere queryable.
   • Led three document-intelligence tools to production — invoice, email, and timesheet parsers — past 90% field-level accuracy on documents that often arrive as photographs of paper.
   • Exposed the agents as async FastAPI services so other teams could integrate without touching a model, deployed via Docker and CI/CD on AWS.

2. ML Solutions Engineer

   Aug 2018 – Sep 2021

   ISSD Bilişim Elektronik A.Ş. · Turkey · Hybrid

   • Shipped real-time incident detection for tunnels and highways with Intel, running optimized models on edge hardware because shipping video to a datacenter was never going to be fast enough.
   • Built a 3D container scanner for port logistics, raising throughput while cutting misreads.
   • Developed train-crossing detection with Huawei, catching hazards at intersections before they became collisions.

• LLM Systems

  • Multi-agent orchestration
  • RAG pipelines
  • Tool calling
  • LangChain
  • Autogen
  • MCP

• ML & Vision

  • NLP
  • Object detection
  • Incident detection
  • Crowd analysis
  • Time series forecasting

• Backend

  • Python (asyncio)
  • FastAPI
  • C++
  • Qt
  • Microservices

• Infrastructure

  • AWS
  • Docker
  • Kubernetes
  • OpenVINO
  • CI/CD

• Data Stores

  • PostgreSQL
  • Qdrant
  • Elasticsearch
  • JanusGraph
  • AWS Neptune
  • Redis
  • MongoDB

1. Incident Detection on Junction Using Image ProcessingarXiv preprint 2104.13437 · 2021
2. Application of the Neural Network Dependability Kit in Real-World EnvironmentsarXiv preprint 2012.09602 · 2020
3. Aiming for Smart Wind EnergyTransactions on Emerging Telecommunications Technologies · 2019